Error Function Calculation Device and Error Function Calculation Method

1. An error function calculation device which performs processing of calculating an error function F M (T) defined by the following equation (1) and having T (positive real number) as a variable, t (0 to 1) as a function variable, and M (M is an integer of 0 or more) as an order, F M ⁡ ( T ) = ∫ 0 1 ⁢ t 2 ⁢ M ⁢ exp ⁡ ( - Tt 2 ) ⁢ ⁢ ⅆ t ( 1 ) the error function calculation device comprising: a coefficient table which stores a value of F M+i (T 0 )/F M (T 0 ) (i is at least one integer among 1 to n, and n is a natural number) calculated for a predetermined T 0 , obtained by an n-th order polynomial approximation of the error function F M (T), or a value obtained by multiplying the value of F M+i (T 0 )/F M (T 0 ) by a constant; and a calculation section which reads the value of F M+i (T 0 )/F M (T 0 ) or the value obtained by multiplying the value of F M+i (T 0 )/F M (T 0 ) by a constant from the coefficient table, and performs at least a part of the processing of calculating a value of the error function F M (T) by a fixed-point arithmetic using the value of F M+i (T 0 )/F M (T 0 ) or the value obtained by multiplying the value of F M+i (T 0 )/F M (T 0 ) by a constant.

2. The error function calculation device as defined in claim 1 , wherein the value of F M+i (T 0 )/F M (T 0 ) or the value obtained by multiplying the value of F M+i (T 0 )/F M (T 0 ) by a constant is a fixed-point value.

3. The error function calculation device as defined in claim 1 , comprising: a decoder which generates an index for reading the value stored in the coefficient table, wherein the decoder generates the index based on an exponent part of the variable T in floating-point representation and a higher-order bit of a mantissa part of the variable T.

4. The error function calculation device as defined in claim 1 , wherein the polynomial approximation is a Taylor approximation.

5. The error function calculation device as defined in claim 1 , wherein the calculation section performs processing of calculating at least a part of the following equation (2) by a fixed-point arithmetic wherein C is a constant and ε is a displacement F M ⁡ ( T ) = ⁢ CF M ⁡ ( T 0 ) ⁢ { 1 + ( - ɛ ) ⁢ { F M + 1 ⁡ ( T 0 ) F M ⁡ ( T 0 ) + ⁢ ( - ɛ ) ⁢ { 1 2 ⁢ F M + 2 ⁡ ( T 0 ) F M ⁡ ( T 0 ) + ( - ɛ ) ⁢ 1 6 ⁢ F M + 3 ⁡ ( T 0 ) F M ⁡ ( T 0 ) + … } } } . ( 2 )

6. The error function calculation device as defined in claim 5 , wherein the coefficient table stores a value of F M (T 0 ) or a value obtained by multiplying the value of F M (T 0 ) by a constant, and wherein the calculation section performs the processing of calculating the value of the error function F M (T) based on a calculation result of a fixed-point arithmetic using the value of F M+i (T 0 )/F M (T 0 ) or the value obtained by multiplying the value of F M+i (T 0 )/F M (T 0 ) by a constant read from the coefficient table and the value of F M (T 0 ) or the value obtained by multiplying the value of F M (T 0 ) by a constant.

7. The error function calculation device as defined in claim 5 , wherein the coefficient table is formed by a memory, storage content of which is rewritten.

8. The error function calculation device as defined in claim 7 , comprising: a transfer control section which transfers data stored in an external memory to the coefficient table, wherein the external memory stores a value of F L (T 0 ) (L is an integer from 0 to M) or a value obtained by multiplying the value of F L (T 0 ) by a constant and a value of F L+i (T 0 )/F L (T 0 ) (i is an integer from 1 to n) or a value obtained by multiplying the value of F L+i (T 0 )/F L (T 0 ) by a constant, wherein the transfer control section reads a value of F N (T 0 ) (N is an integer from 0 to M) or a value obtained by multiplying the value of F N (T 0 ) by a constant and a value of F N+i (T 0 )/F N (T 0 ) or a value obtained by multiplying the value of F N+i (T 0 )/F N (T 0 ) by a constant, and writes the value of F N+i (T 0 )/F N (T 0 ) or the value obtained by multiplying the value of F N+i (T 0 )/F N (T 0 ) by a constant into the coefficient table, the value of F N (T 0 ) or the value obtained by multiplying the value of F N (T 0 ) by a constant and the value of F N+i (T 0 )/F N (T 0 ) or the value obtained by multiplying the value of F N+i (T 0 )/F N (T 0 ) by a constant being stored in the external memory, and wherein the calculation section performs at least a part of the processing of calculating the value of the error function F M (T) by a fixed-point arithmetic using the value of F N+i (T 0 )/F N (T 0 ) or the value obtained by multiplying the value of F N+i (T 0 )/F N (T 0 ) by a constant, the value of F N+i (T 0 )/F N (T 0 ) or the value obtained by multiplying the value of F N+i (T 0 )/F N (T 0 ) by a constant being written into the coefficient table.

9. The error function calculation device as defined in claim 7 , wherein the coefficient table stores a value of F M+n (T 0 ) calculated for a predetermined T 0 , and wherein the calculation section reads the value of F M+n (T 0 ) from the coefficient table, and performs at least a part of the processing of calculating the value of the error function F M (T) by a fixed-point arithmetic using values of F M+n−1 (T 0 ), . . . , and F M (T 0 ) calculated by the following equation (3) and the value of F M+n (T 0 ) from the coefficient table F M - 1 ⁡ ( T ) = 1 2 ⁢ M - 1 ⁢ { exp ⁡ ( - T ) + 2 ⁢ TF M ⁡ ( T ) } . ( 3 )